JPH10257493A - Motion vector detector and image encoder provided with the same - Google Patents

Abstract

PROBLEM TO BE SOLVED: To enable efficient detection of a motion vector by determining the range of search on a reference picture according to prescribed conditions based on the position of the input picture in the reference picture corresponding to a desired image block position to be predicted and determining the final search range by performing correction when the search range is not settled within the reference picture. SOLUTION: When the temporarily set search range is protruded only on the time side in the horizontal direction, a rectangular range determined by (m-i,n-j), (m-i+k,n-j), (m-i+k,n+j) and (m-i,n+j) is protruded out of the reference picture. Therefore, correction is performed for adding the temporarily set search range on the plus side in the horizontal direction. Then, the search of prescribed motion detection is performed in the final search range with the same form and the same area as the search range temporarily set as the rectangular range determined by four points of (m-i+k,n-j), (m+i+k,n+j), (m+i+k,n+j) and (m-j+k,n+j).

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention is used for a motion compensation coding process when coding image data, searches a predetermined search range of a reference screen, and is most similar to an image block to be predicted on an input screen. The present invention relates to a vector detection technique for detecting a position as a motion vector by specifying a range of a motion vector.

[0002]

2. Description of the Related Art In recent years, as a video signal compression method, a compression method by digital processing has been developed. As an example of such a video signal compression method by digital processing, there is a compression method standard called MPEG.
In these compression techniques, a screen is generally divided into small units called macroblocks, and compression is performed on a macroblock basis. According to the MPEG standard, a compression process is performed by dividing one screen into, for example, 45 macroblocks and 30 macroblocks horizontally, with a range of adjacent 16 pixels vertically and 16 pixels horizontally as one macroblock.

The image data divided into macro blocks is
First, motion detection is performed. With motion detection, from the previous screen,
A range similar to the macroblock to be compressed is searched for, and pixel-by-pixel difference data between pixel data of the target macroblock and pixel data of a similar range searched from the immediately preceding screen is obtained. This is a method of increasing the compression efficiency by transmitting a motion vector that is a distance between a similar range and a macroblock. As a result of the motion detection, the difference value of each pixel data becomes a small value, and the variation becomes small. Therefore, it is expected that the effect is improved in the compression by the Huffman coding described later.

[0004] Next, transform coding is performed on the image difference data on which the motion prediction has been performed. Generally, DCT is used for image compression.
(Discrete cosine transform) is often performed. According to the MPEG standard, for example, in the case of a luminance signal, one macroblock is divided into four, and DCT is performed on a range of eight pixels vertically and horizontally. In the result of DCT, coefficients are also arranged in an 8 × 8 matrix. However, if the coefficients are rearranged according to a certain rule, the coefficients can be arranged in order of frequency components.

As described above, by performing transform coding such as DCT, each coefficient is converted into a frequency component. As a feature of the image data at this time, since the signal of the low frequency component is large and the signal of the high frequency component is small, when the components are arranged in the order of the frequency components, the first several coefficients have large values, and are gradually smaller. The data will be arranged so that the values continue. However, in transform coding, it is not possible to reduce the amount of data only by giving a bias to the data. Therefore, quantization and run-length encoding are performed on this data to reduce the amount of data.

[0006] The quantization reduces the precision of data. Specifically, each coefficient is divided by a quantization step value, and the quotient is used as a coefficient. As a result, the coefficient having a small value becomes 0, and the coefficient having a value can be reduced. Generally, when image data is divided into frequency components, the value on the low frequency side is large and the value on the high frequency side is small. Therefore, the coefficient particularly on the high frequency side often becomes zero due to quantization. Therefore, quantization has the meaning of reducing the data amount by cutting high-frequency components.

[0007] Data is biased by the transform coding and quantization, and the number of coefficients having a value of 0 increases. Therefore, these coefficients are compressed by run-length coding. Run-length coding is effective when there are many zero-valued coefficients (zero coefficients) in a coefficient sequence and there are occasional values (non-zero coefficients). , Followed by a non-zero coefficient value (LEVEL). As a result of the quantization, particularly the high-frequency component has a large number of zero coefficients, so that a set of run-length coded coefficients themselves often has a coefficient smaller than the output of the DCT.

After the coefficients are reduced by the run-length coding, the number of bits of each coefficient set is compressed by a variable length code. The variable length code is to reduce the bit amount as a whole by allocating a short bit string to data having a high frequency of occurrence and allocating a long bit string to data having a low frequency of frequency. In the MPEG standard, The correspondence between the data and the bit string is defined in advance. In the variable length code specified by MPEG, a short bit string is assigned when the value is relatively small (LEVEL). Therefore, if the coefficient is only a small value, a short bit string is often assigned. Data volume can be reduced. Also, when the coefficient is large, the total data amount increases. Since the magnitude of this coefficient can be controlled by quantization,
By increasing or decreasing the quantization value according to the tendency of the result of CT,
It is possible to control the amount of output data.
The EG encoding device controls the data amount by this method.

[0009]

As described above, the MPE
In compression techniques such as G, motion detection, irreversible conversion of coefficients by DCT and quantization, and efficient compression of data by using variable-length codes are realized. Due to the use, it is not possible to obtain completely the same information as before compression at the time of reproduction, and an error occurs. This error is the main cause of the deterioration of the image quality of the reproduced image, and it is necessary to minimize the error in order to obtain a reproduced image with good image quality.

In the MPEG standard, a high compression rate is realized mainly by two techniques of motion detection and DCT coefficient quantization. However, in order to improve image quality, that is, to reduce errors in reproduced image data, Some compression due to quantization must be sacrificed. Therefore, in order to improve the compression ratio while maintaining good image quality, it is necessary to improve the ability of motion detection. As described above, motion detection is a method of taking the difference between a compressed block and a reference block and using this difference instead of the pixel data of the compressed block. Since a certain difference data has a small value, it becomes easy to be compressed by the latter technology. In other words, an image coding apparatus having a high image quality and a high compression rate is required to have a higher motion detection capability.

By the way, in the ordinary motion compensation prediction, FIG.
As shown in FIG. 2, assuming that the position of the image block to be predicted on the input screen is a pixel located at the position (m, n) in the horizontal and vertical directions, respectively, In the vertical direction, the closest pixel value is searched for from the search range including the pixel set of ± j, and the motion vector is calculated.

In this case, as shown in FIG. 13, the image block position (m, n) of the input screen to be predicted is located, for example, at the end of the screen, and is mi-m to m + i in the horizontal direction and in the vertical direction. It is also conceivable that the search range determined by the pixel set of n−j to n + j does not fall within the reference screen. For example, FIG.
As shown by a broken line in FIG. 3, assuming that only k protrudes in the horizontal minus direction, actually, m−i + k to m + i in the horizontal direction (m component) and n−j in the vertical direction (n component). The motion vector is calculated using a range determined by a pixel set of ~ n + j (a portion shown by a thick frame in FIG. 13) as a search range.

However, if the image block position of the input screen to be predicted is located at the end of the screen, and the substantial search range in the reference screen is narrowed, in the example of FIG. In spite of the ability to process m ± i pixels in the horizontal direction and n ± j pixels in the vertical direction as the search range, motion vectors can be actually detected only in a narrow search range. Efficiency gets worse. This is one of the problems to be improved in the above-mentioned situation in which the image coding apparatus with high image quality and high compression rate is required to have better motion detection capability.

SUMMARY OF THE INVENTION The present invention has been made to solve the above-described problem, and it is possible to efficiently detect a motion vector regardless of the position of a predicted image block on an input screen. The purpose is to be.

[0015]

Means for Solving the Problems and Effects of the Invention The invention according to claim 1 for achieving this object is used for motion compensated predictive coding, and searches a predetermined search range of a reference screen. A motion vector detecting device that detects a position as a motion vector by specifying a range most similar to an image block to be predicted on an input screen, wherein a search range on the reference screen is an image block to be predicted on the input screen. A search range determining means for determining in accordance with a predetermined condition based on a position in the reference screen corresponding to the position, and, if the search range determined by the search range determining means does not fit within the reference screen, the search A search range that is not present is added to a search range that is within the reference screen and originally fits within the reference screen,
And a search range correcting means for setting a final search range.

According to the motion vector detecting apparatus, a position is detected as a motion vector by searching a predetermined search range on the reference screen and specifying a range most similar to the image block to be predicted on the input screen. First, the search range determining means determines a search range on the reference screen based on a position on the reference screen corresponding to a position of an image block to be predicted on the input screen according to a predetermined condition. The predetermined condition is appropriately set according to, for example, the content of the image to be encoded. Note that this condition does not basically change while processing is performed on the image to be encoded.

If the search range determined by the search range determining means does not fit within the reference screen, the search range correcting means sets the search range that does not fit into the reference screen and within the reference screen. The final search range is added to the search range originally contained.

When the image block position of the input screen to be predicted is located at the end of the screen, the substantial search range in the reference screen is reduced in the conventional apparatus. For example, in the example of FIG. 12, m ± i in the horizontal direction and n in the vertical direction.
In spite of the ability to process ± j pixels as a search range, motion vectors can be actually detected only in a narrow search range, resulting in poor motion vector detection efficiency.

On the other hand, according to the motion vector detecting device, if the search range determined by the search range determining means does not fit within the reference screen, the search range correcting means sets the search range that does not fit. The minute is added to the search range within the reference screen and originally within the reference screen, and is set as a final search range. Therefore, the search ability can be effectively used, and the efficiency of motion compensation prediction is improved. This is very effective in consideration of the fact that image coding with high image quality and high compression rate requires better motion detection capability.

Various correction methods can be considered by the search range correction means. For example, a search range that does not fit in the reference screen is added to a search range that originally fits in the reference screen with the same shape, and the same shape as the search range determined by the search range determination unit is secured. The final search range can be used.Also, a search range that does not fit in the reference screen is added to a search range originally having the same area but changing shape and originally fit in the reference screen. It can also be a simple search range.

As in the former case, even when a search range that does not fit in the reference screen is added with the same shape, the area does not change. In other words, regardless of the method used, if the search range after correction is the same as the search range before correction, that is, if the search range is determined by the search range determining means, the search capability is considered to be kept the same for the time being. Can be Therefore, it can be said that it is preferable to perform the correction so as to have the same area.

However, if a search range that does not fit in the reference screen is added with the same shape as in the former case, the search range after correction has the same shape as the search range determined by the search range determination means. In many cases, it is more preferable from the viewpoint of improving the search ability.

As a method of adding a search range that does not fit in the reference screen but has the same area but a different shape as in the latter case, for example, even if the search range extends in the horizontal direction, Could be added to the side. Of course, it may be added in both the horizontal direction and the vertical direction.

As an image encoding apparatus provided with such a motion vector detecting device, a configuration as described in claim 4 can be considered. That is, an image encoding device that converts an input image signal into pixel data of a rectangular range formed as a set of pixels, and that can output at least motion-compressed and compressed digital data, at least the rectangular range Image storage means for temporarily storing pixel data included in the image data, reference image storage means for holding at least one screen of image data to be referenced for motion compensation, and connection to the image storage means and the reference image storage means The image data in a predetermined range is taken out from the reference image storage means, compared with the pixel data stored in the image storage means, a difference value for each pixel is calculated, and output as pixel difference data. A motion compensating unit, an encoding unit that encodes pixel difference data output from the motion compensating unit, and generates output data; Image reproducing means connected to the image forming means, for reproducing the reference image and storing the reference image in the reference image storing means, and connected to the image storing means and the reference image storing means and stored in the reference image storing means. Motion detection means for searching for image data having the same position as the pixel data obtained from the image storage means, and outputting the position as a motion vector to the reference image storage means. The detecting means is configured as the above-described motion vector detecting device.

As described above, the motion detecting means in this case can effectively utilize its search capability and improve the efficiency of motion compensation prediction as described above as a motion vector detecting device, so that high image quality and high image quality can be obtained. An image encoding device having a compression ratio can be realized. What is set forth in claim 5 is:
In the above-described encoding device, the encoding unit performs transformation encoding on the pixel difference data output from the motion compensation unit and outputs the result, and the encoded transformation data output from the transformation encoding unit. And a variable-length encoding unit that converts the output from the conversion encoding unit into a variable-length code and outputs the output. The image reproduction unit performs inverse conversion on the output from the conversion encoding unit to generate reproduced pixel data. It is characterized by having.

In this case, the pixel data is reproduced by the inverse quantization means and the inverse orthogonal transform means, and this becomes equal to the reproduced image at the time of decoding the compressed data. Therefore, if the motion prediction means creates difference data from the pixel data in the rectangular range output from the data conversion means using the pixel data as reference data, errors such as quantization can be corrected. Output, the amount of data can be reduced compared to the output of raw data.

The above-mentioned transform coding means includes:
For example, it can be composed of an orthogonal transformation unit and a quantization unit, but for example, only the orthogonal transformation unit may be used. In general, DCT (Discrete Cosine Transform) is often adopted as the orthogonal transform by the orthogonal transform means, but other orthogonal transform such as discrete Fourier transform (DF)
T) or the like may be employed. Further, as the variable-length encoding means, it is conceivable to compress the transformed encoded data output from the transform encoding means by, for example, run-length encoding, and further convert the data into a variable-length code.

[0028]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the present invention will be described below with reference to the drawings. FIG. 1 is a block diagram showing a schematic configuration of an image coding apparatus conforming to the MPEG standard provided with a motion vector detecting apparatus according to an embodiment of the present invention. The image encoding apparatus receives a black-and-white image signal conforming to the NTSC system as an input, performs a compression process on the signal, and outputs the result as a compressed stream. This compressed stream has a structure similar to the MPEG standard.

As shown in FIG. 1, the present image encoding apparatus comprises:
An input image signal is received, stored as digital data in pixel units, and pixel data is output in macroblock units (corresponding to a “rectangular range” of the present invention) in a range of 16 × 16 pixels. Storage means 1 for
And pixel data output from the image storage means 1;
The motion compensation unit 2 receives pixel data of an image immediately before output from a reference image storage unit 5 described later, calculates difference data for each pixel, and outputs the difference data. Encoding means 3 for performing encoding for compression, outputting the result to the outside as a compressed stream, and outputting the encoded data to an image reproducing means 4 to be described later for internal reproduction;
Image reproducing means 4 for reproducing pixel data based on the coded data received from the image reproducing means 4; receiving a reproduced image from the image reproducing means 4;
And a reference image storage unit 5 for supplying necessary pixel data to the reference image storage unit 1. The pixel data is received from the image storage unit 1 in units of macroblocks, and the pixel data of the immediately preceding image included in the corresponding search range is stored in the reference image storage unit. And a motion detecting means 6 for searching a range most similar to the macroblock referred to from No. 5 and outputting a motion vector as a result to the reference image storing means 5.

The internal structure and basic operation of each of the above-mentioned units 1 to 6 will be described with reference to FIGS. First, the image storage means 1 includes an image memory 11 and a delay circuit 12, as shown in FIG. The image signal, which is an external input, is an interlaced signal having 525 scanning lines composed of two fields, a first field and a second field. For predetermined 240 scanning lines respectively from the field and the second field, 720 pixel data are taken out from each scanning line and taken into the image memory 11. The pixel data is 8-bit data per pixel, and indicates the luminance of an image.

The image memory 11 can take in one frame of image. The pixel data read and stored in the image memory 11 is rearranged and read for each macroblock in units of 16 pixels in length and width, and is output to the motion detection means 6 as a target of motion detection.

The pixel data is also captured by the delay circuit 12, where it is delayed for a certain period and output to the motion compensating means 2 as a base image for motion compensation. The delay period corresponds to a period from when the motion detecting unit 6 receives the basic pixel data for motion detection to when the motion vector is determined and the pixel data in the range indicated by the motion vector is output to the motion compensating unit 2.

The motion compensating means 2 comprises a subtraction circuit 21 as shown in FIG. 3, and includes motion compensation basic pixel data obtained from the image storing means 1 and motion compensating pixel data obtained from the reference image storing means 5. The reference pixel data is input to the subtraction circuit 21 and the value (Y) obtained by subtracting the motion compensation reference pixel data (B) from the motion compensation base pixel data (A) is output to the encoding unit 3 as “pixel difference data”. I do. Note that the operation (Y =
AB) is performed for each pixel data.

The encoding means 3, as shown in FIG.
A DCT transform circuit 31, a quantization circuit 32, a run-length encoding circuit 33, and a Huffman encoding circuit 34 are provided. Then, the pixel difference data output from the motion compensating means 2 is fetched in macroblock units, and the DCT conversion circuit 31 and the quantization circuit 32 perform DCT conversion and quantization to reduce the data amount. The output from the quantization circuit 32 is “coefficient data” for internal image reproduction.
Is output to the image reproducing means 4. At the same time, the output from the quantization circuit 32 is output to the run-length encoding circuit 3
3 and through the Huffman encoding circuit 34,
It is output to the outside as a “compressed stream”.

As shown in FIG. 7, the image reproducing means 4 includes an inverse quantization circuit 41 and an IDCT (inverse DCT) conversion circuit 42. The coefficient data fetched from the encoding means 3 as an input is inversely quantized by an inverse quantization circuit 41, and further subjected to IDCT conversion by an IDCT conversion circuit 42 to obtain reproduced pixel data. The output from the image reproducing means 4 is this reproduced pixel data, which is output to the reference image storing means 5.

The reference image storage means 5 is a storage means for holding the immediately preceding image for reference, and as shown in FIG. 8, a reference memory 51, an input selector 52, an output selector 53 and an output selector 53. A selector 54 is provided.
The “reproduced pixel data” from the image reproducing means 4 is fetched via the input selector 52, and is input to the corresponding reference memory 5.
Write to 1. The reference memory 51 is a memory having a capacity of 512 kilobytes and has a two-bank configuration of [bank 1] and [bank 2], and writing can be performed in one of the banks. The arrangement of each pixel on the memory conforms to the arrangement on the memory in the image memory 11 described with reference to FIG. The pixel data of the immediately preceding image is stored in the pixel data of the unwritten bank, and the output is performed from here.

The output from the reference image storage means 5 is as follows:
There are two types, “motion compensation reference pixel data” output via the output selector 53 and “motion detection reference pixel data” output via the output selector 54. The reference pixel data for motion detection outputs pixel data in the order determined by the motion detection means 6. The motion compensation pixel data is calculated from the motion vector output from the motion detection means 6 and the address of the base macroblock, and the pixel data in the same range as the macroblock is determined in a predetermined order, that is, the top of the macroblock. Rows are moved in the horizontal direction, and this is sequentially repeated in the next row in the vertical direction.

As shown in FIG. 7, the motion detecting means 6 fetches the motion detection base pixel data output from the image storage means 1 and uses the motion detection base pixel data as a reference to store the reference image storage means 5. , Pixel data of the immediately preceding image is read out as reference pixel data for motion detection, and a pixel range similar to the base pixel data for motion detection is searched. The similarity is determined, for example, by the sum of absolute values of the difference values for each pixel.

The search range on the reference screen is basically shown in FIG. 13 when the position of the macroblock of the basic pixel data on the screen is indicated in the horizontal and vertical coordinate formats (m, n). As described above, the range extended by ± i pixels vertically and ± j pixels horizontally is defined as the search range. That is, the four corners of the rectangular range are (mi, nj), (m + i, n-
j), (mi, n + j), and (m + i, n + j).

However, there is a possibility that the search range set in this way may extend out of the reference screen itself as shown in FIG. In this embodiment, in this case, a predetermined correction process is performed to correct the search range. The details will be described later, and the description of the configuration will be continued.

The internal structure of the motion detecting means 6 is as follows.
As shown in FIG. 7, a basic pixel memory 61 and a motion vector calculating circuit 62 are provided. First, “motion detecting basic pixel data” output from the image storage unit 1 is sequentially stored in the basic pixel memory for each macroblock range. 61 is stored. The motion vector calculation circuit 62 reads “motion detection reference pixel data” from the immediately preceding image stored in the reference image storage unit 5, and compares this with the pixel data of the base pixel memory 61 by the motion vector calculation circuit 62. Then, a vector to the most similar range is obtained as a motion vector. Then, the “motion vector” is output to the reference image storage unit 5.

The above is a description of the internal structure and basic operation of each of the means 1 to 6. The feature of the present invention is that the position (m) of the macroblock of the base pixel data on the screen for calculating the motion vector is calculated. , N), the search range is set so as not to protrude from the reference screen. This can be easily understood by referring to the processing operation related to the determination of the search range in the motion vector calculation circuit 62 of the motion detection means 6 as a specific operation, and the processing will be described with reference to the flowchart shown in FIG. In this sense, the motion vector calculation circuit 62 also corresponds to “search range determination means” and “search range correction means”.

The flowchart of FIG.
This is performed for each "motion-detection base pixel data" output from the. In the first step S10, the position of the motion detection base pixel data (m,
n). At S20, a search range on the reference screen is provisionally set based on the image block position (m, n) to be predicted recognized at S10. The temporarily set search range is a range extended by ± i pixels vertically and ± j pixels horizontally from the image block position (m, n) to be predicted. That is, the four corners of the rectangular range are (mi, n-
j), (m + i, n-j), (mi, n + j), (m
+ I, n + j).

In subsequent S30, it is determined whether the search range provisionally set in S20 falls within the reference screen. Specifically, comparison of the maximum value mmax of the reference screen in the horizontal direction (m direction) with (m + i), comparison of the minimum value mmin with (mi), maximum value nmax of the reference screen in the vertical direction (n direction) And (n + j) and minimum value nmin
And (n-j).

If the search range temporarily set in S20 falls within the reference screen (S30: YES), the process proceeds to S40.
At S20, the search range provisionally set in S20 is determined as the final search range, and the present processing routine ends. On the other hand, S2
If the search range temporarily set at 0 is not within the reference screen (S30: NO), the process proceeds to S50, where the search range is corrected so as to be within the reference screen, and at S60, the corrected search range is set. Is determined as the final search range, and this processing routine ends. Thus, S40 or S6
A search for predetermined motion detection is performed in the search range determined at 0.

Here, the correction of the search range in S50 will be described. As shown in FIG. 9, when the temporarily set search range protrudes by k on the negative side in the horizontal direction (m component) as shown by a broken line, the range shown in FIG.
That is, (mi, nj), (mi + k, nj),
A rectangular range defined by (mi + k, n + j) and (mi, n + j) protrudes from the reference screen. Therefore, the correction in this case is performed by adding a range having the same shape as the range in FIG. 9 to the plus side in the horizontal direction (m component) of the temporarily set search range. Therefore, the final search range is (m−i + k,
n−j), (m + i + k, n−j), (m + i + k, n)
+ J) and (mi-k, n + j). This has the same shape and the same area as the temporarily set search range.

Of course, FIG. 9 shows an example in which the provisionally set search range protrudes from the reference screen and protrudes to the negative side in the horizontal direction (m component). However, there is a case where it protrudes to the plus side, that is, (m + i) becomes larger than mmax. Similarly, there may be a case where the projection extends in both the plus and minus directions in the vertical direction (n component). These are corrected in the same manner as described with reference to FIG. 9, that is, if they protrude to the minus side, they have the same shape (that is, the same area) as the protruding range (see FIG. 9) (see FIG. 9). ) May be added to the plus side, and if it protrudes to the plus side, it may be added to the minus side.

As shown in FIG. 10, the horizontal direction (m
Component and the vertical direction (n-component) may be simultaneously projected. In this case, however, a range having the same shape as the extended range may be added to the inside of the reference screen.
Specifically, considering a case in which the image protrudes by k on the negative side in the horizontal direction (m component) and protrudes by 1 on the positive side in the vertical direction (n component), the search range after correction is (m−
i + k, n-j-l), (m + i + k, n-j-l),
(M + i + k, n + j-l), (mi-k, n + j-
1) The rectangular range is determined by the four points. As a matter of course, this has the same shape and the same area as the provisionally set search range.

As described above, according to the image coding apparatus of the present embodiment, when detecting a motion vector, the search range temporarily set based on the position (m, n) of the image block to be predicted is included in the reference screen. If the search range does not fit, the search range that does not fit is added to the search range within the reference screen and originally within the reference screen, and correction is performed to make the final search range. Therefore, the search ability can be effectively used, and the efficiency of motion compensation prediction is improved. This is very effective in consideration of the fact that image coding with high image quality and high compression rate requires better motion detection capability.

As described above, the present invention is not limited to the above embodiments, and can be implemented in various forms without departing from the gist of the present invention. For example, in the above embodiment, as the correction of the search range in S50 of FIG. 8, the search range that does not fit in the reference screen is added to the search range that originally fits in the reference screen with the same shape, S
Although the same shape as the search range provisionally set in 20 is secured as the final search range, the search range that does not fit in the reference screen originally fits in the reference screen with the same area but a different shape. The final search range can be added to the search range. For example, as shown in FIG. 11, in the case where the image protrudes in the horizontal direction (m component) (see range), it can be added to the vertical direction (n component) side (range, see). However, in this case, the range +
It is necessary to adjust the area of the range to be the same as the area of the range.

As in the above embodiment, even when a search range that does not fit in the reference screen is added with the same shape, the area does not change. That is, no matter what method is used, if the search range after the correction is the same as the area before the correction, that is, the same area as the temporarily set search range, it is considered that the search capability is kept the same for the time being. Therefore, it can be said that it is preferable to perform the correction so as to have the same area.

However, as in the former case, if a search range that does not fit in the reference screen is added with the same shape, the search range after correction has the same shape as the temporarily set search range. It is considered that it is often more preferable from the viewpoint of improvement. In the above embodiment, DC
Although the transform coding means is constituted by the T transform circuit 31 and the quantization circuit 32, for example, only the DCT transform circuit 31 may be used. However, since the compression effect is enhanced by the quantization by the quantization circuit 32, it is preferable to use the DCT conversion circuit 31 as a set. The DCT transform circuit 31 is an example of the orthogonal transform means.
Other orthogonal transforms (discrete cosine transform), for example, a discrete Fourier transform (DFT) may be adopted.

[Brief description of the drawings]

FIG. 1 is a block diagram of an image encoding apparatus for image compression according to an embodiment of the present invention.

FIG. 2 is a configuration diagram of an image storage unit according to the embodiment.

FIG. 3 is a configuration diagram of a motion compensation unit in the embodiment.

FIG. 4 is a configuration diagram of an encoding unit in the embodiment.

FIG. 5 is a configuration diagram of an image reproducing unit according to the embodiment.

FIG. 6 is a configuration diagram of a reference image storage unit in the embodiment.

FIG. 7 is a configuration diagram of a motion detection unit in the embodiment.

FIG. 8 is a flowchart showing a search range determining process in the motion detecting means.

FIG. 9 is an explanatory diagram showing search range correction when an image block position of an input screen to be predicted is located at an end of the screen.

FIG. 10 is an explanatory diagram showing search range correction when an image block position of an input screen to be predicted is located at an edge of the screen and protrudes in both the horizontal and vertical directions.

FIG. 11 is an explanatory diagram showing another method of search range correction when an image block position of an input screen to be predicted is located at an end of the screen.

FIG. 12 is an explanatory diagram showing a search range on a reference screen with respect to an image block position of an input screen to be predicted.

FIG. 13 is an explanatory diagram showing a conventional search range when an image block position of an input screen to be predicted is located at an end of the screen.

[Explanation of symbols]

DESCRIPTION OF SYMBOLS 1 ... Image storage means 2 ... Motion compensation means 3 ... Encoding means 4 ... Image reproduction means 5 ... Reference image storage means 6 ... Motion detection means 10 ... Image memory 11 ... Delay circuit 21 ... Operation circuit 31 ... DCT conversion circuit 32 ... Quantizer 33 Run-length encoder 34 Huffman encoder 41 Inverse quantizer 42 IDCT converter 51 Reference memory 52 Input selector 53 Output selector 54 Output selector 61 Base pixel memory 62 Motion vector calculation circuit

Claims (5)

[Claims] 1. A method for detecting a predetermined search range on a reference screen, specifying a range most similar to an image block to be predicted on an input screen, and detecting the position as a motion vector. A search range determining means for determining a search range in the reference screen based on a position in the reference screen corresponding to an image block position to be predicted in the input screen, according to a predetermined condition; If the search range determined by the search range determination means does not fit within the reference screen, the search range that does not fit is replaced by the search range that is within the reference screen and originally fits within the reference screen. And a search range correcting means for adding a final search range to the search range. 2. The search range correction unit adds a search range that does not fit in the reference screen to a search range that originally fits in the reference screen while keeping the same shape. 2. The motion vector detecting device according to claim 1, wherein the same shape as the search range determined by the above is secured as a final search range. 3. The search range correction means adds a search range that does not fit in the reference screen to a search range originally having the same area but changing shape and that fits in the reference screen. The motion vector detecting device according to claim 1, wherein the motion vector is set as a general search range. 4. An image coding apparatus capable of converting an input image signal into pixel data in a rectangular range formed as a set of pixels, and outputting the digital data as compressed digital data after at least performing motion compensation. Image storage means for temporarily storing pixel data included in the rectangular range; and at least one image data referred to for motion compensation.
A reference image storage unit for holding a screen, and a pixel which is connected to the image storage unit and the reference image storage unit, extracts image data of a predetermined range from the reference image storage unit, and stores the pixel data in the image storage unit. A motion compensation unit that calculates a difference value for each pixel in comparison with the data and outputs the pixel difference data as pixel difference data; and an encoding unit that encodes the pixel difference data output from the motion compensation unit and generates output data. Means, connected to the encoding means, an image reproducing means for reproducing an image for reference and storing it in the reference image storage means; and a reference image connected to the image storage means and the reference image storage means. The image data stored in the storage means is searched, and a range most similar to the pixel data obtained from the image storage means is specified. An image encoding device, comprising: a motion detection unit that outputs to a reference image storage unit, wherein the motion detection unit is configured as the motion vector detection device according to claim 1. . 5. The transform encoding means for transform-encoding the pixel difference data output from the motion compensation means and outputting the result, and the transform code output from the transform encoding means. Variable-length encoding means for converting the encoded data into a variable-length code and outputting the converted data, wherein the image reproducing means performs inverse transform on the output from the transform encoding means to generate reproduced pixel data. The image encoding device according to claim 4, further comprising a transform encoding unit.